Predicting protein properties with spatial statistical machine learning

The aim of the project is to develop algorithms able to construct models for predicting protein function based on the structure, sequence, phylogenetic, interaction information, information about gene expression and evolutionary conservation of structural motifs.

For this purpose we develop new statistical algorithms-relational machine learning utilizing the spatial nature of the tasks using specialized heuristics dependent on the position of evaluating logic formulas and their approximate integration through subspace protein configurations in 3D.